This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
[PATCH]: Fix assumption in gdb.base/savedregs.exp
- From: "David S. Miller" <davem at davemloft dot net>
- To: gdb-patches at sources dot redhat dot com
- Date: Thu, 06 Apr 2006 18:34:25 -0700 (PDT)
- Subject: [PATCH]: Fix assumption in gdb.base/savedregs.exp
This testcase assumes that the location of the dummy frame will
match up to a symbol and source file, so it uses the regexp:
"Stack frame at .* in .*"
But this will not be true for platforms which implement dummy frames
on the stack, like 32-bit Sparc does. The PC will be somewhere on the
stack and this will not resolve to a function name and source file.
Ok to apply?
PS: It is possible to implement 32-bit Sparc dummy calls as an
AT_ENTRY_POINT style dummy call just like 64-bit sparc. I
have an implementation sitting around somewhere, and the only
difficult bit is handling the "if func returns struct, skip
callers unimp instruction" anomaly of the 32-bit Sparc calling
conventions. I plan to dig that up at some point in time.
2006-04-06 David S. Miller <davem@sunset.davemloft.net>
* gdb.base/savedregs.exp: Do not require that the dummy
frame location match up to a symbol and source file location.
--- gdb.base/savedregs.exp.~1~ 2004-11-07 08:47:57.000000000 -0800
+++ gdb.base/savedregs.exp 2006-04-06 18:29:18.000000000 -0700
@@ -81,7 +81,7 @@
dummy {
# Dummy frame's do not have saved registers, and do
# not print <dummy>.
- set pat "Stack frame at .* in .*"
+ set pat "Stack frame at .*"
}
sigtramp {
# Sigtramp frames don't yet print <signal trampoline>.